Loading WHODrug Type C Format In TMS 4.5. - DBMS Consulting, Inc

Loading WHODrug Type C Format In TMS 4.5. Decision Points and Considerations

September 2004

An CompanyElaine Luscombe, i3 Data Services Sunil G. Singh, DBMS Consulting

DBMS Consulting, Inc. © 2004

Oracle Clinical Users’ Group Meeting – September 2004

Acknowledgements

Thanks to Thanks to the OCUG and TMS Focus Group for this opportunity to present this topic

Thanks to Kim Renjdrup and Andy Alasso for their comments and feedback about possible implementations of WHOD T C i TMS

DBMS Consulting, Inc

WHODrug Type C in TMS

Thanks to Daniel von Sydow and Malin Nord at the WHO Uppsala Monitoring Center for their comments into this topicUppsala Monitoring Center for their comments into this topic

Many thanks to the OCUG Presentation committee and Lori Venable for their patience with the completion of theseVenable for their patience with the completion of these presentations.

DBMS Consulting, Inc. © 2004 Presented by: Sunil G. Singh and Elaine S. Luscombe


Goals and Agenda

Explain challenges to loading the WHO Drug Type C Format in TMS 4 5Format in TMS 4.5

Identify the key decision points that must be addressed


before loading

Provide suggestions for possible loading and configurationProvide suggestions for possible loading and configurationoptions

f Provide suggestions on possible future enhancements andchanges both to the WHODrug Type C format and TMS



Brief TMS Overview

TMS (Thesaurus Management System) is normallyintegrated with Oracle Clinical as well as otherintegrated with Oracle Clinical as well as otherapplications.

D i th B t h V lid ti th V b ti TDBMS Consulting, Inc

During the Batch Validation process, the Verbatim Termsare compared to the dictionary data. The Level of thedictionary for which the Verbatim Term is compared isy pknown as the Classification Level.

If an exact match is found then this matched term and If an exact match is found, then this matched term, andother associated data in the dictionary, is returned toOracle Clinical



Brief TMS Overview

TMS defines an exact match as :

Equivalent regardless of the case of the terms

DBMS Consulting, Incand

Equivalent regardless of the number of spacesbetween words of a multiple word term.p



Brief TMS Overview

The process of resolving Verbatim Terms ( VT) which donot exactly match some term in a medical dictionary isnot exactly match some term in a medical dictionary iscommonly called “coding.” In TMS terminology, this iscalled Omission Management.

DBMS Consulting, Inc When a VT is assigned to a dictionary term, the term andits auxiliary data, also called Level Details, is returned toy , ,Oracle Clinical in the next Batch Validation.

Other parts of the medical dictionary known as Derivable Other parts of the medical dictionary, known as DerivableLevels, can be returned to Oracle Clinical at the same time.



TMS Limitations

TMS must have unique terms in the Classification Level ofa dictionarya dictionary.

TMS l t di ti d t b k t O l Cli i lDBMS Consulting, Inc

TMS can only return dictionary data back to Oracle Clinicalif the data exists in a single derivable path. This meansthat for every term in the Classification Level of ayDictionary, TMS must have a single default term from anylevel higher than the classification level in order to returnthat data to Oracle Clinicalthat data to Oracle Clinical.



Overview of WHODrug

The WHODrug dictionaries are published by the UppsalaMonitoring Centre (UMC) on behalf of the World HealthOrganization (WHO).

DBMS Consulting, Inc The WHODrug dictionary represents a collection of allmanufactured drugs that have been reported to them.manufactured drugs that have been reported to them.

The WHODrug dictionary has had four major formats since The WHODrug dictionary has had four major formats since1968, with the current formats of Type B2 and Type C.



Overview of WHODrug

The WHODrug dictionary contains many levels including :

Drug Names Anatomical Therapeutic Classifications (ATCs)


Ingredients Source Codes

M f t Manufacturers

There are abo t 1200+ ATCs there are c rrentl more thanThere are about 1200+ ATCs, there are currently more than60,000 actual unique Drugs in the Type C format.



Differences in the WHODrug Type C Affecting TMS The Drug names themselves are not unique in Type C.

Type C Affecting TMS

ATC codes are now associated to every Drug Name inType C

DBMS Consulting, Inc A Pharmaceutical Product level, which contains the

Pharmaceutical Form (PF), was introduced in Type C

All ingredients and their amounts were introduced in Type C

Th M di i l P d t ID (MP ID) hi h t 7 d The Medicinal Product ID (MP ID), which represents 7 drugattributes, now uniquely identifies a drug (Drug Name,Name Specifier, Country, Manufacturer, All Ingredients w/St th d U it Ph ti l F D C d


Strengths and Units, Pharmaceutical Form Drug Code(DrgRecNum+Seq1+Seq2))


WHODrug Loading Considerations

Since the drug name is not unique in Type C, the drugname alone can not be loaded as the Classification level inname alone can not be loaded as the Classification level inTMS. Therefore, the drug names must be made uniquesomehow.


In making drug names unique in Type C, the TMS built-ing g q ypautomatic matching would potentially be diminished. Someconsiderations have to be made for preserving TMS autoencoder efficiencyencoder efficiency.



Choices

Option 1: Use the Medicinal Product ID itself to make the DrugNames unique in the classification levelNames unique in the classification level

Option 2: Use the logical expansion of the Medicinal Product


p g pID to make the Drug Names unique in the classification leveland possibly populate a VTA Level with Drug Names only Drug NameDrug Name Name Specifier Country Manufacturer Manufacturer All concatenated Ingredients with their Strengths and Units, Pharmaceutical Form Drug Code (DrgRecNum+Seq1+Seq2)


Drug Code (DrgRecNum+Seq1+Seq2)


Choices (cont.)

Option 3: Add an additional level to store the Drug Names only as part of a Classification Group in the TMS WHODrugonly as part of a Classification Group in the TMS WHODrug structure




Option 1 : Medicinal Product IDat Classification Level

Advantages Easy to load

at Classification Level

Easy to load

Disadvantages


g Auto encoding would not be possible

Coders would not have information needed to selectCoders would not have information needed to selectcorrect VTA

SSome have suggested adding the MP_ID to only non-uniquedrug terms. This still leaves many terms (10,000+) which willnot auto encode.


not auto encode.


How it Looks to the Coder


DrugName MP_ID



What it Really Means !



MP_ID 247 is Ampicillin Sodium


What it Really Means !


MP ID 275 is Ampicillin Note the different


MP_ID 275 is Ampicillin Trihydrate.

Note the different Drug Codes


Option 2 : Use the Logical Expansionof the MP ID

Advantages

of the MP ID

Information is available for coders to select appropriate VTA

DBMS Consulting, IncDisadvantages

Nothing auto encodes Nothing auto encodes Load script is more complicated and takes longer Field length may require >CHAR 300Field length may require >CHAR 300



The Logical Expansion of the MP ID


Note the Aspirin Cardio includes the strength

Note the Drug Codes are the


gand form


same


So Nothing Auto Encodes ? Ever ?

Answer - Load Verbatim Term Assignments (VTAs)(VTAs)

This also allows coders to use the filter buttons in TMS O i i M t t h th VTA


TMS Omission Management to choose the VTA Level and only code on the Drug Names if desired

Problem - Over 10,000 drug names are not unique.

Do you have to manually code all 10,000+?

Yes and No !


Yes and No !


Yes - Manually Code All Duplicate Drug Names

Advantages

Duplicate Drug Names

Control of the codes – you can select certain drugs fromspecific countries, or manufacturers, or ingredients, etc.


You could reload same VTAs, once they are selected tonew versions of the dictionary.



Yes - Manually Code All Duplicate Drug Names

Disadvantages

Duplicate Drug Names

As each version is released, you will need to repeat thisexercise.

DBMS Consulting, Inc How long will it take your team to code 10,000+ terms?

S f th t ill i t d b t Some of these terms you will never see in a study, butyou will spend a lot of time on them initially.



No - Don’t do it Manually

Load only the Unique Drug Names as verbatim terms.

Code the others as they are encountered as verbatimterms


terms.

AdvantagesAdvantages Over 40,000 will be able to have VTAs loaded

You only spend time on those you need



No - Don’t do it Manually

Disadvantages

You may give up consistency in decision making if this isdone over time

DBMS Consulting, Inc Many of the most common drugs encountered are in thisgroupgroup

You need to repeat this with each new version of thedi idictionary



Alternative “No” - do it Systematically

AdvantagesS f f Same script can be used for each new version of thedictionary

Logic can be applied that is consistent across all termDBMS Consulting, Inc

Logic can be applied that is consistent across all termchoices

The script will run faster than your team can do theThe script will run faster than your team can do thework!



Decisions That Need to be Made

How do you decide which drugs should have VTAs?W f l th t ll th d h i th We feel that all the drugs having the same DrgRecNum and Seq1 and can have a VTA selected.


selected. The same DrgRecNum and Seq 1 mean the drug is

the same drug with the same Preferred term and the same ingredients. Further, WHODrug will continue to support the

D R N d S b ( th d tDrgRecNum and Seq numbers (see the document titled The New C Format: New Features that accompanies each version of the dictionary).


accompanies each version of the dictionary).SS2

SS2 (can we reference a specific document from WHO-UMC here?)Sunil Singh, 7/20/2004


Alternative “No” - do it Systematically

Disadvantages

Decisions still need to be made on the logic to be used


Some terms will not have VTAs because the same drugname by different countries/manufacturers are reallydifferent drugsdifferent drugs

TMS loading Script development is complex and willTMS loading Script development is complex and will take some time to run!

This does not account for PF or strength.



Why not Assign Global VTAs to DrugNames with Multiple Drug Recnrs ?

In WHODrug Type, 4th Quarter 2003, there were about 775cases where there were Drug Names with multiple Drug

Names with Multiple Drug Recnrs ?

cases where there were Drug Names with multiple DrugRecord numbers, and therefore potentially differentingredients, Preferred Names, and ATCs.


There are many reasons why this has happened :



Reasons for Multiple Record Numbers

A strategic decision by a manufacturer to change the activeingredients to improve the product over time but keep theingredients to improve the product over time, but keep thesame Drug Name due to market share and brandrecognition


The lack of availability of some active ingredients in somecountries or geographies, including cases where the rawmaterials are not available or are banned by a country forhuman use or importp



Reasons for Multiple Record Numbers

The purchase or acquisition of one company or brand byanother combined with a strategic decision to keep theanother combined with a strategic decision to keep thesame brand recognition and market share purposes, but toalso change or improve the drug which might change the

ti i di tDBMS Consulting, Inc

active ingredients

The lack of enforcement of intellectual property rights orpatents in some countries, where the same Drug Name isused illegally and manufactured with completely differentg y p yingredients. WHO-UMC is still obligated to report thecreation and use of these drugs.



Example of Same Drug Name in Different Countries

Benadryl in Italy

Different Countries




Example of Same Drug Name in Different Countries

Benadryl in the United Kingdom

Different Countries




Changes in the Drug

Robitussin AC




Default Algorithm for Preserving TMSAuto Coding WHODrug Type C

1. Where only one Medicinal Product ID exists for a specificDrug Name, then make this Drug Name a Verbatim Term.

Auto Coding WHODrug Type C

2. Where multiple Medicinal Product IDs exist for a specificDrug Name but only one Medicinal Product ID has


Drug Name, but only one Medicinal Product ID hasSeq2='001' (i.e., a unique Preferred Term), then make thisDrug Name a Verbatim Term.

3. Where multiple Medicinal Product IDs exist for a specificDrug Name, and multiple Medicinal Product IDs haveSeq2='001' but only one Medicinal Product ID hasCOUNTRY='UNS' (i.e., the Country code is Unspecified),then make this Drug Name a Verbatim Term.


then make this Drug Name a Verbatim Term.


Default Algorithm For Preserving TMSAuto Coding WHODrug Type C

4. Where multiple Medicinal Product IDs exist for a specificDrug Name, and multiple MP IDs have Seq2='001' and


Drug Name, and multiple MP IDs have Seq2 001 andmultiple MP_IDs have COUNTRY='UNS' , then make theDrug Name of the lowest Medicinal Product ID theVerbatim Term


Note: If you use the Maximum MP_ID your VTAs willbe altered in newer versions of the dictionary asbe altered in newer versions of the dictionary asMP_IDs are added.



Default Algorithm for Preserving TMSAuto Coding WHODrug Type C

5. Where multiple Medicinal Product IDs exist for a specificDrug Name and no Medicinal Product ID has Seq2='001'


Drug Name, and no Medicinal Product ID has Seq2 001but only one Medicinal Product ID has COUNTRY='UNS'(i.e., the Country code is Unspecified), then make thisD N th V b ti T


Drug Name the Verbatim Term.

6. Where multiple Medicinal Product IDs exist for a specificp pDrug Name, and no Medicinal Product ID has Seq2='001'and no Medicinal Product ID has COUNTRY='UNS', thenmake the Drug Name with the lowest Medicinal Product IDmake the Drug Name with the lowest Medicinal Product IDthe Verbatim Term.



Decision Tree for Selecting VTA

Identify Drugs With The Same DrgRecNum & Seq1Num

- Select Where One Drug Has SeqNum2=001 (Preferred Name) As VTA #1

-If >1 Drug Has SeqNum2 =001

- Select Where The Country = Unspecified (UNS)


q ( )-If >1 Country = UNS Select Lowest/Highest MP ID As VTA

- Select Where The Country = Unspecified - For All Drugs Where Seq1≠ 001-Select Where The Country = unspecified#2

y p(UNS)- If >1 Country = UNS Select Lowest/Highest MP ID As VTA

- For All Drugs Where Country ≠ Unspecified

- Select Lowest/Highest MP ID As VTA#3



View of TMS, Classification Level inWHODrug Type CWHODrug Type C




Derivable Level Details with Drug Name and Preferred Drug Name in WHODrug Type C




Want One More Option ?

Option 3: Creating a new level in TMS between Classificationlevel and VTlevel and VT.

During the OPA 5.0 Enhancement meeting from 11-12 May


2004, Oracle OPA Development has suggested creating anintermediate level within the TMS dictionary itself to preserveauto coding.g



Option 3 : New Level in TMS

Consider the possibilities…..

Load Classification level with Drugname MPID Load Classification level with Drugname_MPID This level is now unique and since your are not going to try to

“code” anything to it, it does not need additional description

DBMS Consulting, Inc For all drugs at the classification level, determine if there

are multiple drug names and, for those, if they have thediff t D R N d S N 1same or different DrgRecNum and SeqN 1

If the Drug Name is unique, load it at the middle levelintermediateintermediate




If the Drug Name is not Unique, but all theDrgRecNum|SeqN1’s are the same verify that the PFDrgRecNum|SeqN1’s are the same, verify that the PF,Strength and Ingredients are all the same and then chooseone to DrugName_MPID to represent the rest and load that


at the middle intermediate level.

You can use the logic presented earlier or some otheri il d ibl l i f l ti hi h illsimilar, reproducible logic for selecting which one will

represent all these drugs that are the same.

This would be a stored relation in the TMS dictionary This would be a stored relation in the TMS dictionarystructure




Additionally, some other extensions of this option might make it more useful:make it more useful:• Where an entry is made the Intermediate Middle Level, also

make a corresponding entry to a VTA level (combining


p g y ( gaspects of Option 2)

• All other non-unique drug names including multiple cases of the same drug name with different DrgRecNums| Seq1 canthe same drug name with different DrgRecNums| Seq1 can be loaded to the new mid-level with Drugname_PharmaceuticalForm_Strength_ Ingredients

• This level will now give you the information your coders need to make decisions on the correct one to chose for coding.

• A primary link could be considered between the Unique Drug


p a y cou d be co s de ed bet ee t e U que ugclassification level and the Intermediate Level.


Other Considerations of Option 3: Loading New Level

Loading an additional dictionary level would require structural changes to the existing relations between the

Loading New Level

structural changes to the existing relations between the Unique Drug Names classification level and the intermediate Middle Level in order to change the default Drug Name/MP ID, that is, the default coding behavior.


Drug Name/MP_ID, that is, the default coding behavior. These changes would be made through Maintain Repository Data, which can lead to other data-related mistakes or exposures of bugs which can be very difficult tomistakes or exposures of bugs which can be very difficult to repair.

Some thinking from a validation perspective might be that th t t d d t f d di ti h ldthe structure and data of a vendor dictionary should never be modified. In this context, modification of VTAs is not the same as modification of the WHO-UMC’s dictionary data or di ti t t


dictionary structure


Dictionary Updates and Reducing Data ScopeData Scope One consideration is whether or not all of the Drug data

should be loaded. Why not parse all of the Drug Names only and simple load these Drug Names?only and simple load these Drug Names?• Not Loading the MP_ID or loss of the MP_ID will make

updating this dictionary very difficult. This is because the default TMS APIs for updating the dictionary


default TMS APIs for updating the dictionary, TMS_LOAD_DICTIONARY.MigrateRelations and TMS_LOAD_DICTIONARY.MigrateTerms expect a unique DICT CONTENT CODE in the dictionary which comes fromDICT_CONTENT_CODE in the dictionary which comes from the vendor which can be compared with queries against the vendor source data to determine what DICT CONTENT CODEs to insert/update/delete._ _ p

• Additionally, during the dictionary load process, it is not required to specify a DICT_CONTENT_CODE nor is uniqueness enforced! But during update calls using the


uniqueness enforced! But during update calls using the TMS_LOAD_DICTIONARY API, it is a de facto expectation.


Dictionary Updates and Reducing Data Scope (2)Data Scope (2) This means not having the MP_ID for all of the WHODrug

source data will make updating very difficult. Calls to sou ce data a e updat g e y d cu t Ca s toTMS_USER_MT_DICTIONARY for updating, inserting and deleting terms will have to be made on a separate basis, without the benefit of the TMS migration APIs


without the benefit of the TMS migration APIs. Additionally, if only part of the drug data is loaded (a

reduction in the data scope), it may be possible to make a p ), y pvalidation argument that the dictionary loaded in TMS was not actually a representation of the WHO-UMCs WHODrug dictionary but a customized dictionary which is adictionary, but a customized dictionary which is a proprietary to a single organization, which may introduce some additional validation requirements.



Some Suggested TMS EnhancementsEnhancements Not Requiring Classification Level to be Unique, but instead

requiring the combination of the TERM_UPPER along with LEVEL_1 to be unique?

Allowing configurable column lengths in TMS DICT CONTENTS for TERM and TERM UPPER so


TMS_DICT_CONTENTS for TERM and TERM_UPPER, so that customers which require a longer Term field (such as Option 2) can extend their classification content past 300 Characters?Characters?

Create a TMS codelist which is default enables which requires dictionaries to be loaded with DICT CONTENT CODE hi h t ll d i ?DICT_CONTENT_CODEs which are not null and unique?



Additional Discussions

1. Consider the slide for Aspirin - Should medications becoded to the level of detail available in the dictionary?coded to the level of detail available in the dictionary?

>If so, how will you accomplish this?

2 C ld d b b ilt i t OC/TMS th t ll fDBMS Consulting, Inc

2. Could a procedure be built into OC/TMS that allows formerging specified fields together and have THAT be theVTA that is classified in TMS?

3. Do your CRFs include all the needed information?

4. Will the information be used?



Questions / Contacts

S il G Si hSunil G. SinghDirector, DBMS Consultinge-mail: [email protected]: +1 888 463 4751


Pager: +1 888 463 4751Mobile: +1 860 983 5848

Elaine LuscombeManager, Data Services Supporte-mail: [email protected]@Phone: +1 973 348 1288

Visit us at our respective booths in the Exhibit Hall


Loading WHODrug Type C Format In TMS 4.5. - DBMS Consulting, Inc

Documents

Transcript of Loading WHODrug Type C Format In TMS 4.5. - DBMS Consulting, Inc